ACR Meeting Abstracts

ACR Meeting Abstracts

Abstract Number: L11

Rheumatology Diagnostics Utilizing Artificial Intelligence (ANA Reader©) for ANA Pattern Identification and Titer Quantification

May Choi1, Farbod Moghaddam1, Mohammad Sajadi1, Ann E. Clarke1, Sasha Bernatsky2, Karen Costenbader3, Irene Chen4, Murray Urowitz5, John Hanly6, Caroline Gordon7, Sang-Cheol Bae8, Juanita Romero-Diaz9, Jorge Sanchez-Guerrero10, Daniel Wallace11, David Isenberg12, Anisur Rahman13, Joan Merrill14, Paul Fortin15, Dafna Gladman16, Ian Bruce17, Michelle Petri18, Ellen Ginzler19, Mary Anne Dooley20, Rosalind Ramsey-Goldman21, Susan Manzi22, Andreas Jönsen23, Graciela Alarcón24, Ronald Van Vollenhoven25, Cynthia Aranow26, Meggan Mackay26, Guillermo Ruiz-Irastorza27, S. Sam Lim28, Murat Inanç29, Kenneth Kalunian30, Soren Jacobsen31, Christine Peschken32, Diane Kamen33, Anca Askanase34, Marvin Fritzler35 and Mina Aminghafari1, 1University of Calgary, Calgary, AB, Canada, 2Research Institute of the McGill University Health Centre, Montreal, QC, Canada, 3Brigham and Women's Hospital/Harvard Medical School, Boston, MA, 4UC Berkeley and UCSF, Berkeley, CA, 5Self employed, Toronto, ON, Canada, 6Dalhousie University, Halifax, NS, Canada, 7University of Birmingham, Birmingham, United Kingdom, 8Hanyang University Medical Center, Seoul, South Korea, 9The National Institute of Medical Sciences and Nutrition, Mexico City, Mexico, 10Krembil Research Institute, Toronto, ON, Canada, 11Cedars Sinai Medical Center, Studio City, CA, 12Department of Ageing, Rheumatology and Regenerative Medicine, Division of Medicine, University College London, London, United Kingdom, 13University College London, London, United Kingdom, 14Oklahoma Medical Research Foundation, Oklahoma City, OK, 15Centre ARThrite - CHU de Québec - UniversitéLaval, Quebec, QC, Canada, 16University of Toronto, Toronto Western Hospital, Toronto, ON, Canada, 17Queen's University Belfast, Belfast, United Kingdom, 18Johns Hopkins University School of Medicine, Baltimore, MD, 19SUNY Downstate Health Sciences University, New York, NY, 20UNC physician network, Chapel Hill, NC, 21Northwestern University, Chicago, IL, 22Allegheny Health Network, Pittsburgh, PA, 23Lund University, Lund, Sweden, 24The University of Alabama at Birmingham, Oakland, CA, 25Amsterdam UMC, Amsterdam, Netherlands, 26Feinstein Institutes for Medical Research, New York, NY, 27Biobizkaia Health Research Institute, Bilbao, Spain, 28Emory University, Atlanta, GA, 29Istanbul University, Istanbul, Turkey, 30UC San Diego, La Jolla, CA, 31Rigshospitalet, Copenhagen, Denmark, 32University of Manitoba, Winnipeg, MB, Canada, 33Medical University of South Carolina, Johns Island, SC, 34Columbia University Medical Center, New York, NY, 35Mitogen Diagnostics Corp, Calgary, AB, Canada

Meeting: ACR Convergence 2024

Date of first publication: October 24, 2024

Keywords: , , , , ,

Session Information

Date: Monday, November 18, 2024

Title: (L01–L14) Late-Breaking Posters

Session Type: Poster Session C

Session Time: 10:30AM-12:30PM

Background/Purpose: Antinuclear antibody (ANA) immunofluorescence (IFA) patterns and titers are a key part of rheumatology diagnostics, however, there is considerable intra- and inter-laboratory variability with manual interpretation. Replacing manual interpretation with a standardized and automated approach could help reduce variability, increasing laboratory accuracy and efficiency. We developed machine learning (ML) models to aid laboratories in ANA pattern and titer interpretation, including a model for the nuclear dense fine-speckled (DFS) ANA pattern (AC-2), a rare pattern among systemic autoimmune rheumatic disease (SARD) patients that decreases the likelihood of these conditions.

Methods: 13,671 ANA images from SLE patients enrolled in the Systemic Lupus International Collaborating Clinics Inception Cohort (SLICC, n=2,825 images), non-SLE subjects enrolled in the Ontario Health Study (OHS, n=10,639 images), and the International Consensus on ANA Patterns (ICAP, n=207 images) were analyzed. All SLICC and OHS ANA were performed in one central laboratory using IFA on HEp-2 cells (NovaLite, Werfen, SD) and read on a digital IFA microscope (NovaView, Werfen, SD). As the reference standard, one laboratory technologist (HH) with >30 years of experience in ANA studies interpreted ANA patterns and titers for each image. We developed and compared the performance of eight ML models for 13 ANA pattern recognition. Four convolutional neural network (CNN) models and an image feature extractor were developed to differentiate AC-2 from AC-4 (speckled) and AC-30 (nuclear speckled with mitotic plate staining) patterns, which appear similar but are associated with SARDs. To evaluate ANA titer, we used an ML technique for imaging processing that identified individual HEp-2 cells in the ANA images and then calculated the cell illuminance and cut-offs corresponding to each titer (1:80-1:5120). Fifty images were randomly selected to compare the titer classification based on image processing with the lab technologist as the reference standard.

Results: We identified one ML model with the best performance for ANA pattern identification compared to the reference with a high area under the curve (AUC) score of 83.4% and more modest performance in other metrics (Figure 1). For the AC-2 models, all four CNNs performed similarly with high AUC scores (96.5%-97.1%) (Table 1). There was a strong correlation between titers reported by the identified model and the technologist’s interpretation (Spearman rank 0.93, p < 0.0001), where the titers reported were identical or differed by only one dilution in most cases (96.0%).

Conclusion: ML has the potential to become a highly effective and efficient approach to evaluating ANA patterns and titers. The performance of our model is expected to improve as we continue to train our models with more images. The AC-2 model can already discriminate the nuclear DFS pattern from other SARD-related patterns with a high degree of accuracy, potentially speeding up the differentiation of those at risk vs. not at risk of SARDs. Future external validation studies and ML models to predict more complex and multiple ANA patterns and titers are underway.

Supporting image 1

Figure 1. Area-under-the-curve (AUC) scores for the thirteen anti-nuclear antibody (ANA) patterns using the ANA Reader© model, which had the best performance compared to seven other machine learning techniques. It had the highest area under the curve (AUC) score of 83.4%, with modest weighted accuracy of 68.4%, precision of 67.1%, sensitivity of 70.1%, and F1 score of 67.2%. The ANA patterns with the best performance were centromere (AUC 0.97) and pleomorphic patterns (AUC 0.97). On average, there were five images per patient sample for SLICC, three images per patient sample for OHS, and one image per patient from ICAP. In total, there were 512 patients in the SLICC cohort, 3,559 individuals in the OHS cohort, and 207 patients from ICAP who were included in the study. Among the 13,671 ANA images, 6,307 images contained at least one ANA pattern (>=1:80) from SLICC (n=2,806 images), OHS (n=3339 images), and ICAP (n=162 images). 80% of the images were used for model training and the remaining 20% for validation.

Supporting image 2

Disclosures: M. Choi: AstraZeneca, 2, Celltrion, 2, GlaxoSmithKlein(GSK), 2, Mallinckrodt, 2, MitogenDx, 2, Organon, 2, Werfen, 2; F. Moghaddam: None; M. Sajadi: None; A. Clarke: AstraZeneca, 2, 5, 7, Bristol-Myers Squibb(BMS), 2, GlaxoSmithKlein(GSK), 2, 5, 7, Otsuka Pharmaceutical, 1, Roche, 1; S. Bernatsky: None; K. Costenbader: Astra Zeneca, 2, Exagen, 5, Gilead, 5, Merck, 5, Neutrolis, 2, 10; I. Chen: None; M. Urowitz: AstraZeneca, 2, GlaxoSmithKline, 2, 5, 7, Lilly, 7, UCB, 2; J. Hanly: None; C. Gordon: Astra-Zeneca, 2, 7, Centre for Disease Control, 2, 7, MGP, 2, 7, Sanofi, 2, 7, UCB, 2, 5, 7; S. Bae: None; J. Romero-Diaz: None; J. Sanchez-Guerrero: None; D. Wallace: AstraZeneca, 2, 7, Aurunia, 2, 7, Eli Lilly and Company, 2, 7, EMD Serono, 2, GlaxoSmithKline, 2, 7; D. Isenberg: AstraZeneca, 2, Eli Lilly, 2, GSK, 2, Merck Serono, 2, Novartis, 2, Servier, 2, UCB, 2; A. Rahman: Lilly, 7; J. Merrill: AbbVie, 2, 12, Paid instructor, Amgen, 2, AstraZeneca, 2, 5, Aurinia, 2, Biogen, 2, 12, Paid instructor, BMS, 2, 5, 6, 12, Paid instructor, Eli Lilly, 2, EMD Serono, 2, Genentech, 2, GSK, 2, 5, Kezar, 2, Merck, 2, Pfizer, 2, Provention, 2, Remegen, 2, 12, Paid instructor, Sanofi, 2, Takeda, 2, Tenet, 2, UCB, 2, Zenas, 2, 6, 12, Paid instructor; P. Fortin: AbbVie, 1, AstraZeneca, 1, Lilly, 1; D. Gladman: AbbVie, 2, 5, Amgen, 2, 5, Bristol Myers Squibb, 2, 5, Celgene, 2, 5, Eli Lilly, 2, 5, Galapagos, 2, 5, Gilead, 2, 5, Janssen, 2, 5, Novartis, 2, 5, Pfizer, 2, 5, UCB, 2, 5; I. Bruce: AstraZenaca, 2, 5, 6, Dragonfly Therapeutics, 2, GSK, 2, 6, Janssen, 2, 5, 6, Lilly, 2, Takeda, 2; M. Petri: Amgen, 2, AnaptysBio, 2, Annexon Bio, 2, Arthros-FocusMedEd, 6, AstraZeneca, 2, 5, 6, Atara Biosciences, 2, Aurinia, 2, 5, 6, Autolus, 2, Bain Capital, 2, Baobab Therapeutics, 2, Biocryst, 2, Biogen, 2, Boxer Capital, 2, Cabaletta Bio, 2, Caribou Biosciences, 2, CTI Clinical Trial and Consulting Services, 2, CVS Health, 2, DualityBio, 2, Eli Lilly, 2, 5, EMD Serono, 2, Emergent Biosolutions, 2, Escient Pharmaceuticals, 2, Exagen, 5, Exo Therapeutics, 2, Gentibio, 2, GSK, 2, 5, iCell Gene Therapeutics, 2, Innovaderm Research, 2, IQVIA, 2, Janssen, 5, Kezar Life Sciences, 2, Kira Pharmaceuticals, 2, Nexstone Immunology, 2, Nimbus Lakshmi, 2, Novartis, 2, Ono Pharma, 2, PPD Development, 2, Proviant, 2, Regeneron, 2, Seismic Therapeutic, 2, Senti Biosciences, 2, Sinomab Biosciences, 2, Steritas, 2, Takeda, 2, Tenet Medicines, 2, TG Therapeutics, 2, UCB, 2, Variant Bio, 2, Worldwide Clinical Trials, 2, Zydus, 2; E. Ginzler: None; M. Dooley: None; R. Ramsey-Goldman: None; S. Manzi: Astra Zeneca, 2, 5, Cugene, 2, Eli Lilly, 2, Exagen, 2, 5, 10, GSK, 2, UCB, 2; A. Jönsen: None; G. Alarcón: None; R. Van Vollenhoven: AbbVie, 2, 7, AstraZeneca, 2, Biogen, 2, Biotest, 2, Bristol Myers Squibb, 2, 5, Eli Lilly, 5, Galapagos, 2, 7, Gilead, 2, GlaxoSmithKline, 2, 7, Janssen, 2, 7, Pfizer, 2, 7, 12, Support for educational programs; institutional grants, Roche, 12, Support for educational programs; institutional grants, Sanofi, 2, Servier, 2, UCB, 2, 5, 7, Vielabio, 2; C. Aranow: GlaxoSmithKline, 2, 5; M. Mackay: None; G. Ruiz-Irastorza: None; S. Lim: ACR, 4, AstraZeneca, 5, Bristol Myers Squibb, 5, GlaxoSmithKline, 2, Pfizer, 2, UCB, 2; M. Inanç: None; K. Kalunian: AbbVie/Abbott, 2, Alliance, 2, Amgen, 2, AstraZeneca, 2, Aurinia, 2, Biogen, 2, Bristol-Myers Squibb(BMS), 2, Eli Lilly, 2, Equillium, 2, Genentech/Roche, 2, Gilead, 2, Janssen, 2, Lupus Research, 5, Nektar, 2, Pfizer, 5, Sanford Consortium, 5, Vielabio, 2; S. Jacobsen: None; C. Peschken: AstraZeneca, 2, Eli Lilly, 2, GlaxoSmithKlein(GSK), 2; D. Kamen: None; A. Askanase: AbbVie/Abbott, 1, Amgen, 1, AstraZeneca, 1, 5, Aurinia, 2, Eli Lilly, 5, GlaxoSmithKlein(GSK), 2, 5, Idorsia, 5, Pfizer, 5; M. Fritzler: Mitogen Diagnostics Corporation, 8, 11, 12, Medical Director, Werfen, 1, 2, 7; M. Aminghafari: None.

To cite this abstract in AMA style:

Choi M, Moghaddam F, Sajadi M, Clarke A, Bernatsky S, Costenbader K, Chen I, Urowitz M, Hanly J, Gordon C, Bae S, Romero-Diaz J, Sanchez-Guerrero J, Wallace D, Isenberg D, Rahman A, Merrill J, Fortin P, Gladman D, Bruce I, Petri M, Ginzler E, Dooley M, Ramsey-Goldman R, Manzi S, Jönsen A, Alarcón G, Van Vollenhoven R, Aranow C, Mackay M, Ruiz-Irastorza G, Lim S, Inanç M, Kalunian K, Jacobsen S, Peschken C, Kamen D, Askanase A, Fritzler M, Aminghafari M. Rheumatology Diagnostics Utilizing Artificial Intelligence (ANA Reader©) for ANA Pattern Identification and Titer Quantification [abstract]. Arthritis Rheumatol. 2024; 76 (suppl 9). https://acrabstracts.org/abstract/rheumatology-diagnostics-utilizing-artificial-intelligence-ana-reader-for-ana-pattern-identification-and-titer-quantification/. Accessed .

« Back to ACR Convergence 2024

ACR Meeting Abstracts - https://acrabstracts.org/abstract/rheumatology-diagnostics-utilizing-artificial-intelligence-ana-reader-for-ana-pattern-identification-and-titer-quantification/